STORM TRACK: March 31, 1983 (Volume 6 Issue 3)

Back Index Fwd

Tornado Sound Experiences

By David Hoadley

(Once again, Storm Track breaks new ground as it presents for the first time in print anywhere to the best of our knowledge, a tornado sound study based on documented personal experiences of trained/experienced observers.)

The common perception of most people about tornadoes is of a large funnel shaped cloud extending from a black sky, roaring like an express train. Over time and with much photographic study, storm researchers have found widely varying tornado shapes (symmetrical columns, tubes and ragged, swirling cloud masses), including occasionally the absence of any condensation column at all. Also, the normal cloud base color was found to vary from dark blue-gray to reddish/brown (dust laden) gray. In hundreds of storm encounters, the Editor has experienced only one storm where the rain-free cloud base turned almost black (deep black-gray with firey red breaks) at mid-afternoon. But what about tornado sounds? How much actual advance warning do they give, how often, with what size storms, etc?

To answer these questions, Storm Track sent a one page questionnaire to the more active chasers and received fourteen detailed responses (from about 20 inquiries), including 113 reports on 91 different tornadoes, one funnel, four gustnadoes and one roaring CB hole. When my 1979 data on the Hurricane David tornado in Fairfax County is added (also derived from other questionnaires), then a total of 126 reports on 92 tornadoes from 26 respondents will be considered. This is the data base for the study.

(Of course, the Editor is quick to point out that this data is based on memories and the varying quality of on-scene notes from each chaser (some storm reports are 20 years old). Consequently, it may contain some errors and even an overall bias to report only storms that were sound producers or large and close enough to be potential sounders. On the other hand, this group of respondents represents the best trained and experienced tornado chasers in the world. Consequently, their reports are based on skills of observation and attention to detail not normally found among the lay public.)

Several decisions were made in reviewing the data to resolve inconsistencies: (1) Three funnel cloud reports were received, including one sound producer. However, since the questionnaire specified "Tornado Sounds," all but the tornadic sounding funnel were excluded to avoid biasing the data with vortex reports which could not be documented by F-scale damage. The one report, however, is included and so identified in the text. (2) Where an F-scale damage estimate was given as two numbers (e.g. "2-3" or "1-2"), the higher number was used. (3) Where a range of distance from the tornado was given (e.g. 10-15 miles), the difference was split (e.g. 12.0 miles). (4) Where time reports appeared inconsistent from different chasers at the same storm, those reports from the most experienced chasers were used. In such cases, an initial count of 4 or 5 different tornadoes might become 2 or 3, after correcting for aberrant time reports.

It is now useful to highlight some of the data conclusions, before looking at the more detailed analyses. Noteworthy is that out of 92 tornadoes, one funnel and four gustnadoes, only 14 produced sound reports. Of this number, the average threshold distance for sound from tornadoes is 1.5 miles. Some of the larger tornadoes do announce themselves at a greater distance. The maximum sound/distance report of four miles was from the Binger, Oklahoma storm of May 22, 1981. If the five Binger sound reports are backed out, the average thresh-hold distance drops to slightly less than a mile (most other storms being smaller than the Binger one). If the average threshold distance of 2.6 miles for the five Binger reports is added as one report, the overall distance increases only slightly to 1.1 miles for all other single reports.

On the other hand, big storms don't always announce thsmselves. The F4 Union City tornado was not noted for its roar. Charles Vlcek notes that this tornado was "quiet. I read an account of a man who popped his head out of a storm cellar too soon (he didn't hear anything) and almost got it knocked off -- the twister was practically next door." Also, the powerful Ames, Iowa tornado (1/2 mile wide) of June 13, 1976 was notable at some close-in locations for the absence of sound -- even at full throttle! Thus, this type of warning can't be relied on even for the super giants, not to mention the more frequent mid-sized storms. Of these F3's, 6 of 7 were observed at less than 2.5 miles with no sound whatsoever, other than normal wind noise through wires or tree limbs. (Of course, the concept of meaningful data on threshold distance to hear tornadic sounds presupposes an even distribution of chasers around and radially from the vortex in all directions, including some who have completely "checked out", and stand directly underneath the wall cloud. However, this is the only data available, so it is used with this assumption, although the actual distribution of chasers is strongly influenced by storm structure with the tendency to congregate along the SE-SW flank.) Now, let's look at some specific data displays which characterize and support these conclusions.

Fig. 10 shows the distribution of all vortex reports within an east-south radius of 7 miles and a west-north radius of 3 miles. Each report is shown as a single number from Fujita's estimate scale for tornadic damage (0, 1, 2, 3, 4 or 5), based on each observers estimate as to storm size/potential for (or actual) damage. Individual groupings of numbers are in no particular order but vere entered as recorded from the data, sometimes along the same axis and distance from the tornado (small circle at upper left center). Therefore, each number represents the location of the observer and the F-scale estimate of the tornado seen from that location. Noteworthy are the maximum ranges for the threshold of sound for different F-scale storms. Observation sites receiving sound are underlined. The Binger tornado of May 22, 1981 projected out to four miles to the southeast. The Altus storm of May 11, 1982 projected south 3 1/2 miles. F1 and F2 storms share the furthest reach for their group at 1 1/2 miles to the southwest and west. However, the uneven distribution of chasers does not assure that sound generating storms sent acoustical signals symmetrically in all directions, to the same distance. Eric Rasmussen noted that the Binger storm "was louder 4 miles to the south than ... to the east."

Figure 10 and 11

Figure 12

Also note, as has already been mentioned, the number of observations reporting no sounds which are well within the maximum range for comparable intensity storms. Charles Vlcek made a prescient observation when he wrote "You probably need to be less than a mile from an F2 tornado to hear anything. Most sounds probably are due to the interaction of wind with objects on the ground (trees, buildings, etc.); if over open country (i.e. a treeless plain), there probably is no low pitched sound whatever." While ST can't verify the terrain features for the no sound reports, John Brown tends to confirm this in treeless Colorado: "So far as I am aware, none of the 14 or so tornadoes observed this year in Colorado (by Boulder or Fort Collins storm chasers) were associated with any distinctive or unusual sounds ..." Moreover, Roy Arnold has done some pioneering research on measurement and conceptual explanations of tornadic sounds. Some excerpts from his article "Storm Acoustics" in Edwin Kessler's three volume series, "Thunderstorms - A Social, Scientific & Technological Docu- mentary": "Study of storm sounds ... has been underexamined, and literature devoted to storm acoustics is relatively sparse. Tornado acoustics ... has received very little analytical attention. ... Major interruptions in fluid flow by ground obstacles, as the base of a funnel moves along the earth's surface, are most likely what produce the principal audible tornado sounds. Eddying of the flow generates sound, and acoustic power output is determined by the flow velocity (Powell, 1964; Lighthill, 1952, 1954). ... Interest in genesis of tornadic sounds ... will be influenced by the supply of sound recordings to the scientific community. ... Propagation effects are also important, ... What is observed at a recording station not only will be a function of the intervening atmosphere but also will depend on the terrain, since sound originates very near the ground. ... Recorded sounds from a tornado are in short supply, but there is encouraging evidence from the one sound analysis reported in the literature that useful information on the tornado itself can be derived from detailed sound study.'

Two different eyewitness accounts tend to confirm Professor Arnold's explanation of the "most likely" cause. Tim Marshall reported the apparent source of "tornadic sound" from a twister 9 miles to the north. From a bluff overlooking the Pease river (with minimum intervening terrain interruptions), a noise was heard, some of which may have occurred after vortex breakdown. "Upon arriving at the tornado path, we saw sheet metal all over, wrapped around telephone poles, lying on the pavement, etc. We found out that the tornado hit a barn about a mile east of the road, which was the source of the debris. Therefore, I believe that the sound was the sheet metal impact instead of the rotating updraft." The most striking account was by Eric Rasmussen for a May 17, 1981 storm only a half mile away: "No sound occurred until the tornado hit two homes. The clackety clack roar was unmistakably associated with the destruction of these two structures." An additional Rasmussen account is also quite interesting for a May 29, 1980 Wellington, Texas encounter: "The first 'touched down' as a small dust, whirl (10 yards in diameter) and moved directly toward Bruce Campbell and myself, passing about 20 yards to our north. A sound like the buzzing of a swarm of bees or the constant discharge of an electrostatic generator was heard coming from the ground, where the tornado was in contact with it."

However, there are also some anomalous reports (although an apparant minority) which don't fit the terrain interface explanation. Space doesn't allow a full recounting, but two earlier Storm Track articles detailed three different encounters with tornadic like roaring sounds from high within CBs. No tornado was present with any of them! The first account by the Editor ST, Vol. 2, No. 6) discussed a May 22, 1968 storm cell about 12-14 miles west of Caldwell, Kansas. About 4 miles west of me was a rainfree CB base with an apparent hole in the middle, rotating cyclonically. Daylight appeared through the hole, with very dark and solid base surrounding it. A powerful, deep roar was heard that continued for several minutes, without oscillation or interruption. The earliest, tornadoes occurred 7 and 18 minutes later and 25 miles NE of this location! The other two reports were written up by Charles A. Doswell, III in a later newsletter (ST, Vol. 4, No. 2). In both instances (May 30, 1976 near Jacksboro, Texas and May 28, 1977 in western Oklahoma), somewhat isolated and dissipating CBs produced roaring sounds without tornadoes. "The storm was of the 'dryline' type ... very small ... appeared to be rotating anticyclonically. ...we heard a steady, muted roaring sound. As nearly as we could tell from a distance of about 3-5 miles to its NE, the sound was coming from the storm's anvil, near where the storm tower joined it from below." The May 28 storm was very similar, except that it rotated cyclonically. "Once again, the sound seemed to originate high up in the anvil, near the point where the storm's tower joined it." Finally, Eric Rasmussen recounts an interesting sound experience with the Alfalfa, Oklahoma tornado of May 22, 1981: "When the tornado was a medium sized columnar condensation funnel to the ground, the sound source seemed to be the entire tornado. However, when it dissipated to a small pointed funnel, the roar was definitely from the tiny funnel. This leads me to wonder if the sound generation is associated with rapid condensation processes." The last word is still to be written on this subject.

Some interesting individual reports round out our eyewitness accounts. Bruce Jensen reports a tornado near Pampa, Texas, "that Eric Rasmussen, Mark Mabey and I witnessed at very close range (possibly ... about 1/4 mile). It was quite large (about 100 yards across) and, while our tape recorder was not working, I recall no sound other than just wind." The "just wind" mentioned by Bruce "was, by the way, about 60 kts at that point, S-SE of the funnel." On May 23, 1981, Roy Britt "heard a very low, faint sound from the wall cloud that produced the Sulphur, Oklahoma tornado. There was no funnel cloud visible, only a wall cloud several miles west of my location. I was told by Randy Zipser and Chuck Robertson that it was a 'roaring hail-shaft.'" Mike Watts also reported a 'roaring hail-shaft' for a May 19, 1977 storm with tornado 2-3 miles to the northwest prior to the roar. Don Burgess reported on another stormy day: "No tornado sound was heard, but a couple of minutes after tornado dissipation, when the wall cloud came overhead, I heard a slight 'swishing' or 'hissing' sound (barely perceptible) associated with rapid cloud base rotation (not a funnel). Mark Mabey recounts a tape recording in the path of a "very low wall cloud in central Oklahoma." A University of Mississippi tape recorder "in a steel case" was deployed by the side of the road, as the prudent chasers moved to a safer position. "... we may have recorded the sounds of a funnel cloud passing directly overhead. The tape starts off with birds singing and not much wind noise. As the tape goes on, the wind starts to increase and finally peaks with the wind blowing very hard. At this time, it sounds like someone is just blowing directly into the microphone. Then this lets up, and the frequency definitely changes. The sounds on the tape now are of a whining, whirling nature. It almost sounds like the swarming of bees ... This lasts for about 20-30 seconds ... and then the hard, blowing wind returns. Shortly after this, we pick it up, and that's it." The Editor reported a similar sound account in ST, Vol. 3, No. 2. (Revised and published in the April, 1981 AMS Bulletin, Vol. 62, No. 4) on a Hurricane David tornado in Fairfax County. At one point, this short lived F2 to F3 tornado passed diagonally (SE to NW) across Whitacre Road, in a residential area of Fairfax County. The resident within a hundred feet to the north of the most heavily damaged home reported that sound "increased in volume to a roar." 500 feet north, it sounded like a freight train roar. 200 feet south it was described as a "steady whoosh." Most striking were the reports from both ends of Whitacre, about 400 feet south and 900 feet north of the heaviest damage. "It hummed, the whole house buzzed" (north end), and "buzzing saw or like a million bees (south end). These reports approximate the central axis and boundaries of the tornado's path. In short, the center roared while the edges buzzed. Eric Rasmussen winds up our potpourri with Binger: It "started very close to our south with the sound of strong winds whooshing, as opposed to roaring. The latter part of Binger produced a loud freight train roar."

Fig. 11 shows the different characteristics of some sound reports, as an enlargement of the data in Fig. 10. Fig. 12 relates the direction of reported tornadic sound ("T") with the location of reported cloud-to-ground lightning. For example, if the tornadic sound comes from the Northwest and the CG lightning comes from the north, a data point, is entered 45 deg to the right of the "T". The three boxes within Fig. 12 show different lightning intensities from bottom to top, or light, moderate and intense. The questionnaire asked the respondent to check only one for each storm. The largest number of data points is shown for the "light" reports. The Figure shows a slight tendency for central grouping of the CG reports along the axis of tornadic sounds, suggesting a possible relationship. However, it should also be noted that no CG lightning was reported in B sound cases and none in 14 other tornado accounts. I have not yet found a meaningful way to relate other data in this study.

Several significant conclusions are suggested by this study, regarding warning lead time and human safety. First, and most importantly, only 35% of tornadoes within maximum possible hearing range produced audible sounds.'

(1) Sound warnings from the largest tornadoes, F0 and greater, can be heard up to four miles away. However, half of the time, you probably have to be within 2 miles to hear (6 of 8, within four miles, were heard). This would provide four minutes warning for a 30 MPH storm; 2 minutes at 60 MPH. Of course, this applies to the optimum circumstance, where you are standing outside (as were most chasers), at the threshold of the audio range. If the family is preparing dinner inside a normal, insulated home with storm windows, exterior sounds could be attenuated by 25% (guesstimate). This would reduce the warning lead time to from 1 1/2 to 3 minutes. While this warning covers only 2-5% of all such, large tornadoes, it does cover a far greater percentage of the fatality producers (Although a search of reference files couldn't document a number, the Editor believes that these super storms account for over 70% of the fatalities).

(2) If an F3 tornado is approaching, the audio threshold for 50% of sound producing storms may be about one mile out-of-doors (Only 1 out of 8, within 3 1/2 miles, was heard). This is a net loss of about 50% lead time over the F4 storm. Our family at the dinner table now only has about 45 seconds to 1 1/2 minutes to seek shelter.

(3) The remaining 80% of the smaller tornadoes, if they produce sound at all, will only be audible within a half to one mile 50% of the time (only 6 of 22, within 1 1/2 miles, were heard). Our family going to bed could have as little as 25 to 45 seconds to do something! Of coarse, they may take dubious comfort, while running for cover, that their chances of surviving the short notice variety are much greater.

In summary, tornado sounds are not reliable warnings end should not be counted on. If you do hear it, you don't have time to put on your slippers, light, your pipe, and stroll outside to look around. You do have time to run to the basement, inner utility room, etc., immediately!

One interesting idea to come out of this: If terrain interface with the vertex is the principal sound generator, then it should be possible to estimate different average lead warning times for different communities. Perhaps a family in a Kansas City suburb can expect a two minute warning, on the average, while a family in Liberal only gets 30 seconds for the same size storm (the approaching terrain being basically flat and featureless). -- one more reason to move to the big city.

Some of the many unanswered questions which remain: (a) Can clouds act as echo chambers to transmit/distort sound? (b) Does the storm structure/vortex in some way organise sound signals? (c) How do sound reports correspond to discernible radar hook echoes? (d) Will sound signatures someday be used with radar signatures to forecast tornadoes? (e) How do we explain roaring CBs without tornadoes? (f) Why are some big tornadoes so quiet? (g) Downbursts are reported to roar too, but only near the ground? (h) Are animals possible indicators for warning sounds, inaudible to human ears? etc. etc.